Senior Full-Stack Python AI Developer – Local Computer Vision & Generative AI Dashboard (MVP)

We are seeking an expert Full-Stack Python and AI Engineer to develop a secure, privacy-focused desktop/tablet web application prototype (MVP). The system processes multi-angle static diagnostic photographs to map physical geometry, analyze surface luminance gradients, calculate local metrics against an internal rules matrix, and generate localized simulation outputs via an offline image-to-image pipeline. This project is fully blueprinted with comprehensive wireframes designed at 1440x1024 resolution. This is a bootstrapped MVP, and we are aiming for a rapid, lightweight, and highly modular build. Core Project Technical Framework -Frontend UI: Built entirely in Streamlit or Gradio (Python-native web layouts) to eliminate heavy frontend framework overhead and accelerate development time. -Facial Mapping: Utilizing MediaPipe Face Mesh to track 468 3D absolute landmarks and map an absolute coordinate system onto static images. -Anatomical Segmentation: Implementation of Meta’s Segment Anything Model 2 (SAM 2) for precise localized region isolation. -Generative AI Pipeline: Local deployment of open-weights models (SDXL or Flux.1) configured via ControlNet Inpainting (Canny/Depth/Tile) to perform localized, high-fidelity texture and shape modifications while strictly preserving subject identity. -Data Analytics Loop: Automated image comparison using Structural Similarity Index (SSIM) and Intersection over Union (IoU) vectors to calculate objective match percentages between simulated models and real follow-up inputs. -Backend Architecture: Written in FastAPI / Python with a PostgreSQL relational database. -Deployment: Containerized and deployed on Serverless GPU Infrastructure (e.g., RunPod, BentoML) to optimize computing costs and maintain idle environments at near $0 overhead. Project Scope, Timeframe & Budget -Duration: Estimated at 3–6 months to complete the foundational, stable core application. -Experience Level Required: Expert Only. This requires low-level machine learning model orchestration and local cloud infrastructure tuning. -Budget Structure: Fixed-Price milestone contracts. The overall listed budget is a placeholder. Please submit your actual proposal based on your estimated hours to complete the MVP milestones, using our pre-built Figma wireframe blueprints. Milestone Breakdown Blueprint: -Milestone 1 (Core Infrastructure): Ingestion of multi-angle static photos, image standardization via OpenCV, and mapping of absolute-anchored interactive dots using the MediaPipe 468 FaceMesh index matrix (supporting click-to-open data tooltips and long-press-to-drag repositioning coordinate snapping). -Milestone 2 (The Local AI Pipeline): Offline orchestration of SAM 2 and SDXL/Flux ControlNet workflows inside localized inpainting masks on an isolated serverless GPU instance to handle regional shape-warping and texture-refining simulations. -Milestone 3 (Analytics, Export, and UI Finish): Integration of the SSIM/IoU verification loop, automated 1:1 image stitching for social media layout exports with built-in metadata mask filters, and completion of the right-side timeline tracking sidebar synced with an inventory/billing ledger state manager. Strict Security & Legal Constraints -Zero External API Dependencies: The application must run completely locally/privately. Outbound data calls to commercial web APIs (such as OpenAI, Anthropic, Midjourney, Replicate, etc.) are strictly prohibited to maintain absolute trade secrecy for upcoming patent filings. -NDA & IP Assignment: Before receiving the technical master blueprint, database schemas, or visual Figma wireframes, the selected top candidates must sign a robust Intellectual Property Assignment and Non-Disclosure Agreement (NDA). All code generated belongs 100% to the client from the moment it is written. Required Applicant Attributes & Qualifications -To be considered for this global role, you must meet the following strict criteria: --Identity & Location Verified: Your Upwork profile must feature the official Upwork identity verification badge --Job Success History: You must maintain a 90% or higher Job Success Score (JSS) on the platform with verified client testimonials in machine learning or computer vision. --Language Fluency: Native or completely Fluent written and spoken English is mandatory. You must be able to clearly discuss architectural constraints and translate clinical protocols into code logic. --Local Model Experience: You must possess verifiable, hands-on experience loading, containerizing, and running raw diffusion weights and ControlNet pipelines entirely locally via the Python diffusers library on Linux/GPU environments. --Coordinate Matrix Expertise: You must be highly proficient with OpenCV matrix operations and understand how to bind custom interactive front-end visual components dynamically to absolute mathematical coordinate systems (like the 468-point MediaPipe mesh index). How to Apply -Please submit a concise proposal detailing your specific experience with local ControlNet inpainting setups, MediaPipe landmark integration, or Streamlit analytics dashboards. -(Please answer in your application cover letter):In our application, a user can manually drag a coordinate marker across the face canvas on a tablet. How would you programmatically ensure that when the user releases the marker, the system instantly snaps it to the closest MediaPipe landmark ID and automatically refreshes that specific localized database record in Python without introducing layout delays?